58577 Community-Based Health Insurance in Lao P.D.R. Understanding Enrollment and Impacts The World Bank, November 2010 Community-Based Health Insurance in Lao P.D.R.: Understanding Enrollment and Impacts World Bank, November 2010 Summary Community Based Health Insurance (CBHI) is one of the four main risk-protection schemes in Laos and is expected to be one of the building blocks to achieving universal coverage in the future. However, after 9 years of pilot projects, coverage remains very low for reasons that are only partially understood. This study was conducted to better understand the factors affecting enrollment in CBHI, members' experiences with the scheme, and the impact of CBHI on members' utilization and out-of-pocket expenditures. A household survey was administered to 1000 CBHI-enrolled households and 2000 comparison households (a total of 14,804 individuals) and a village survey was administered to the village chief in 87 villages. The study found that enrollment in CBHI is influenced by several factors at the household level, including age and education of household head, household size, socioeconomic status, quality perceptions and exposure to CBHI. Health status, risk preferences, having a pregnant woman in the household, and having more women of reproductive age also influence enrollment, confirming that adverse selection (the tendency for households with a sick family member to enroll) is present in the scheme. At the village level, significant determinants of enrollment include: urban/rural mix, the availability of other health care options in the village, age and education of the village chief, exposure to CBHI, village size and the location of the scheme. The impact findings showed that insured individuals have higher utilization rates and lower out-of-pocket expenditures than non-members. CBHI members are also more likely than non-members to use the referral system for their care. Although the findings show positive impacts for CBHI members, few households actually benefit due to the scheme's low coverage. While the study points to programmatic changes that could be made to strengthen CBHI, it also highlights the challenges of scaling up coverage through CBHI and makes some suggestions for complementary approaches to achieving universal coverage. This note was prepared as part of a World Bank program of analytic work on health financing in Lao PDR, and in collaboration with WHO, the London School of Hygiene and Tropical Medicine (LSHTM) and the Ministry of Health (MOH). Core members of the study team included Sarah Alkenbrack (LSHTM), Magnus Lindelow (WB), and Bart Jacobs (Lux-Development, previously WHO). Phetdara Chanthala and Sophavanh Thisty assisted with the study design and implementation arrangements. Field work was implemented by Indochina Research. Helpful comments on the draft versions of the note were received from Bayar Bayarsaikhan, Alexis Bigeard, Genevieve Boyreau, Kara Hanson, Christoph Kurowski, Anne Mills, Jean-Marc Thomé and Adam Wagstaff. The authors would like to thank Dr. Bouaphat Phonvisay (Deputy Chief of CBHI team in Ministry of Health) and her team for their collaboration on this study. Finally, the team is extremely grateful to the interviewees who took the time to participate in this study. The note summarizes findings from the study; further analysis and details will be disseminated in a separate research report. Background private and state-owned enterprises and health equity funds (HEFs) for households living in extreme poverty. Health care in Lao P.D.R. is primarily financed through Together, the four risk-protection schemes are considered direct out-of-pocket payments by households. These the main building blocks of health financing in Lao PDR payments include expenditures on services and drugs from and the government is considering various options for public health facilities, pharmacies, private providers, scaling up coverage to achieve universal coverage by traditional providers, and facilities outside the country. As 2020.1 However, outside the Civil Servants' scheme, in other countries, high reliance on out-of-pocket coverage rates are low, with approximately 1.7% of the financing in Lao PDR forces individuals to either reduce population enrolled in CBHI, 1.5% enrolled in SHI and utilization of health care, ultimately prolonging or 2.1% enrolled in HEFs in 2009 (See Figure 1). worsening health, or make out-of-pocket payments to cover medical-related costs, leading to risk of impoverishment [1, 2]. In an effort to increase health service utilization, decrease health-related out-of-pocket expenditures, improve health outcomes and generate resources for the health sector, the Government of Lao PDR is trying to expand coverage of health insurance and risk protection schemes [3]. CBHI is one of the main risk-protection schemes in the country 1 and targets the informal workforce. Other health A decree is in process to merge all social health protection schemes. While a merging of schemes may result in increased protection schemes operating alongside CBHI are the Civil efficiencies, on its own it is unlikely to increase coverage in the Servants' Scheme (CSS), Social Health Insurance (SHI) for population. 1 Does CBHI membership influence the source of Figure 1. Insurance coverage in Lao PDR health care? SHI Do CBHI households incur lower out-of-pocket CSS 6.3% 1.5% CBHI payments for health care than the uninsured? 1.7% HEF 2.1% Overview of CBHI CBHI has become one of the key risk-protection schemes and is expected to play an important role in helping the country move toward universal coverage in the health Uninsured sector. Introduced in 2001 as a pilot project by the 88.3% Ministry of Health (MOH), the scheme has received technical assistance from WHO and financial support from the United Nations Human Security Fund. Currently, Agence Française de Développement (AFD) supports the MOH with scheme expansion in 2 provinces. The office MOH, 2009; SSO database, SSO Jan 2009; CSS/SASS responsible for managing the scheme is the Health databases, 2009; HEF 2009 annual report, MOH 2010 Insurance Program, within the Department of Planning and Finance, MOH. The MOH contracts hospitals to provide services for CBHI members3 and a gatekeeping Given that CBHI occupies a prominent position on the system requires CBHI members to first seek services at health financing agenda, a study was conducted to better the contracting facility in their district. The benefit understand the status of the CBHI program and identify package for CBHI members is similar to the health care challenges and opportunities to expand enrollment. The benefits in the two formal schemes: it covers outpatient study also examined the impact of the scheme on health and inpatient services including primary health care, care utilization, source of care, and out-of-pocket specialist services, diagnostic tests, and prescribed drugs expenditures.2 that are available at the hospitals.4 The main target group for the CBHI scheme is defined as Key policy questions addressed by the study households who are self-employed or working in the The study addressed the following research questions: informal economy and are not covered by other social protection schemes. This group comprises approximately Enrollment 52 percent of the population.5 Enrollment in the scheme Who is enrolling in CBHI? How are insured takes place at the household level and the cost of households different from the uninsured? premiums varies according to urban or rural residence, What are the most important factors affecting and number of household members. The contribution enrollment? collection rates (see Appendix 1) were originally set at What are the most important reasons for not between 2.5 to 3% of average household income [4]. enrolling in CBHI? However, the contribution rates have not been updated How do households perceive CBHI? What has since 2005, despite average inflation rates of 5.5% per been members' experience with the scheme? year since that time [5]. What percentage of targeted households intends to enroll/maintain enrollment in CBHI in the Household contributions to the scheme are collected on a future? monthly basis by the village collector, who receives LAK 2,000 (US$ 0.25) for each newly enrolled household and Impact LAK 1,000 for each monthly contribution. However, a Are CBHI households more likely to use health recent study noted several problems with the fee care services than uninsured households? collection system: insured households complained that 3 In most districts, the MOH contracts with the district hospital to provide services. However, in Luang Prabang and Vientiane 2 The study was designed and carried out by a team of Province there is no district hospital and the MOH instead researchers from the World Bank and London School of contracts with a provincial hospital. 4 Hygiene and Tropical Medicine (LSHTM), with assistance from The benefit package excludes treatment of injuries, drugs the World Health Organization (WHO). A local research team purchased in pharmacies, and services outside the country. 5 was hired to carry out data collection under the supervision of This figure is an approximation made by taking the total the World Bank/LSHTM team. The study was funded by the population and subtracting the formal sector and their World Bank. dependents, the military, and the poor. 2 collectors stopped collecting fees or did not come on a care workers are subsidized by the Government; all other consistent basis, and village collectors reported that fees costs incurred by CBHI members are expected to be were too low to cover the costs of travel [6]. As a result of covered by the household premiums. Ninety percent of these problems, some villages have moved to a system the amount collected through premiums is paid to the whereby villagers make their payments directly to the contracting hospitals regardless of actual use by CBHI account manager at the district hospital. beneficiaries, and the remaining 10% covers administrative costs. In Vientiane Capital, this capitation To date, CBHI has been implemented mainly in urban and payment amounted to between LAK 40,000 to 45,000 semi-urban areas, but the MOH intends to expand to (US$4.70 - $5.30) per insured person in 2009 [4]. The more remote areas in the future. The rationale for starting capitation is split between the district and the referral in urban/semi-urban areas was that health care services hospitals, according to need. However, in the absence of are of a reasonable quality and the socioeconomic status of additional subsidies, the capitation payments are the target population was deemed high enough that insufficient to cover the cost of services. As a result, households could afford the premiums.6 However, after 9 several central hospitals in Vientiane Capital have refused years of piloting, CBHI in Lao PDR continues to face the to contract with the CBHI scheme [4]. same challenges that most other countries implementing CBHI have experienced. One of the main challenges is the In addition to the problems of low coverage and low coverage rates. Figure 2 shows that rolling out the insufficient funding, high drop-out rates7 and late scheme to new districts has not resulted in substantial payments have been reported as challenges affecting the increases in enrollment. By July 2009, the schemes were scheme. For example, an average of 40 percent of operating in 19 districts but reached only 7% of the members made their payments during the two month population in the targeted districts (and 13% of the warning period that is imposed after payment is due [7]. population in the targeted villages within those districts). Anecdotal evidence also claims that the scheme suffers This is the equivalent to 1.7% of the total population. It is from adverse selection, the phenomenon by which high-risk expected that scale-up of the scheme to more remote individuals (e.g. the chronically ill) have an incentive to areas will pose further challenges due to low population enroll in health insurance at a certain premium when they density, poor geographical access to contracting facilities, have a known need for services [8]. A recent study also and limited acquaintance with modern health care among describes various management challenges of the scheme, ethnic minorities. including insufficient staffing, insufficient technical capacities and scarce financial resources at all levels [9]. Figure 2. Expansion of CBHI, 2002-2009 Study Approach 1,500 Coverage in 2009 The study was designed with reference to the international 1,250 13% of pop. in targeted villages literature on CBHI and previous studies on CBHI in Lao 7% of pop. in targeted districts PDR [6, 10, 11]. The primary method of the study was a Thousands 1,000 1.7% of total population household survey, which was administered from January 750 through April 2009, using a cross-sectional case- 500 comparison design of 1000 CBHI-enrolled households and 2000 comparison households, amounting to a total of 250 14,804 individuals for which information (about 0 individuals and the household) was collected. A village survey was administered to the village chief in the 87 Jun-05 Dec-02 Nov-05 Apr-06 Dec-07 May-03 Sep-06 Feb-07 May-08 Oct-03 Oct-08 Aug-04 Jul-07 Mar-04 Jan-05 Mar-09 villages where the study was conducted.8 Information CBHI enrollees from the two surveys was merged to permit analysis of Pop'n in target villages variables at three levels: individual, household and village. Total pop'n in targeted districts Six focus group discussions with members, non-members Source: CBHI office, Department of Planning and Finance, MOH 7 The drop-out rate at a given time includes both temporary Another challenge facing the CBHI scheme is financial (late payers) and permanent drop-outs. Although drop-outs are sustainability: the scheme is currently not generating reported to be a problem by the CBHI office, it is not possible enough revenues to cover the cost of services and drugs to calculate the drop-out rate because the administrative data offered to CBHI members. Only the salaries of health only includes numbers enrolled and not the numbers joining and leaving the scheme. 8 The MoH approved the implementation of the survey. Ethical 6 Risk-pooling is at the district level but CBHI is not always approval for the study was granted by the ethics committees at rolled out across all villages in a district because some villages the National Institute of Public Health in Laos and at LSHTM in are too remote or do not have good access to the hospital. the United Kingdom. 3 and drop-outs were also conducted to better understand households were selected.10 The response rate for the factors affecting enrollment.9 CBHI and comparison populations was 99.7% and 96.9%, respectively. Data collection Through preliminary focus group discussions, efforts were Data analysis made to understand potential factors affecting enrollment To determine the factors influencing enrollment in CBHI, in CBHI prior to survey design. Survey instruments were probit analysis was undertaken. Results from this analysis then designed to capture self-selection into the scheme. are presented as predicted probabilities, which represent For example, the voluntary nature of the scheme makes it the probability that a household will enroll in CBHI when prone to adverse selection. High-risk characteristics such all other factors are held constant at their mean value. as chronic illnesses, disabilities, or even different attitudes When we compare predicted probabilities for and preferences for health care, have a direct effect on an representative individuals, we can see how the individual's use of health services. Therefore, if these probability of enrollment changes as the variable of variables are not taken into account in the analysis, the interest (e.g. education level) changes [12]. results of the impact evaluation could be misleading. For example, a higher utilization rate among the insured could To measure the impact of insurance, we used a method be the result of a greater need for services (due to more sick known as propensity score matching (PSM) ­ a method family members or a chronic disease in the family), rather that has been used to evaluate the impact of social than improved access due to insurance. Controlling for programs [13], including job training programs [14-17], the differences in these risk and health factors gives a more education programs [18], and more recently, health reliable estimate of the effects of insurance on utilization, insurance schemes [19-23]. PSM uses a range of variables expenditures, and other outcomes. Because perceptions of to construct a single variable known as the propensity illness and poor health differ across households, this score for each observation. The propensity scores study used multiple measures of health status to represent the predicted probability of being enrolled in adequately capture factors affecting enrollment. CBHI and the scores are used to match CBHI observations to comparison observations. This type of matching, which Study location and sample selection is further explained in the technical report, seeks to The study took place in 87 villages across 6 districts from control for observable differences in characteristics Vientiane Capital, Vientiane Province, and Champasak between the CBHI and comparison populations. After Province (See Figure 4). The majority of the sample was matching, the differences in outcomes (e.g. utilization, drawn from urban and semi-urban areas, reflecting the expenditures) between the groups should then represent areas where the majority of the schemes have been the impact of being enrolled in CBHI. The impact implemented to date. evaluation was conducted at the individual level, rather than the household level, given that individuals within a Figure 3. Location of study districts household have different rates of utilization and out-of- pocket expenditures. 11 Study Findings Which households are enrolling in CBHI and how are they different from households without insurance? A comparison of background characteristics between CBHI and comparison households allows us to identify both who is enrolling in CBHI and how insured households differ from the uninsured. Comparisons between the two groups on various household-level 10 CBHI member households were randomly selected from The survey team selected CBHI and comparison member lists in villages. Comparison households were randomly households from villages where CBHI had been in place selected from the village registry. 11 for at least two years. Households were randomly selected The comparison group was constructed using kernel matching, with a Gaussian (normal) kernel and bandwidth of using a two-stage cluster sample approach. For every 0.02. The matching for one outcome was done using CBHI household recruited to the sample, two comparison psmatch2 in Stata and standard errors were bootstrapped with 100 replications. Additional outcomes were estimated using a 9 Further details of the qualitative methodology, analysis and weighted regression, with kernel weights constructed using the results are presented in the full report. propensity score. 4 characteristics (without controlling for other variables) are probability of enrollment increases for those with presented in Appendix 2 and are summarized below: vocational training or post-secondary education. CBHI households are larger, more likely to be Household size. The findings on household size indicate that married, and more educated than non-CBHI households with six or more family members are households. On average, the household head in CBHI significantly more likely to enroll in CBHI. This makes households is older than in comparison households. sense given that the scheme's premium structure provides CBHI households have significantly higher household an incentive for larger households to enroll: as family size consumption levels than comparison households but increases, the cost of the premium per person decreases. similar per capita consumption rates. Among those Although the probability of enrollment is higher for who are employed12, CBHI households are more households with 4 to 5 household members than for 1 to 3 likely than comparison households to hold a long- household members, this difference was not significant. It term contract. is possible that the incentive structure is only strong enough to attract households with at least six family CBHI households are less healthy, have more elderly members. household members, more women of reproductive age and more pregnant women, relative to Health status and risk perceptions/attitudes. The findings show comparison households. CBHI households are also that having worse than average (self-assessed) health does relatively more risk averse (i.e. less likely to take not significantly increase enrollment. However, risks).13 households in which a family member has either a chronic Attitudes toward different sources of care are similar illness or had difficulty performing regular activities in the among CBHI and non-CBHI households. However, past three months (an indicator of illness) were CBHI households report a higher perception of significantly more likely to enroll in CBHI than households quality of health care at the district hospital. with no signs of illness. Moreover, households with CBHI members are more likely to have attended a multiple signs of illness (either within the same family CBHI campaign (See Appendix 3). (However, the member or across family members) were even more likely effectiveness of the campaign in improving awareness to enroll in CBHI than households with just one sign of seems to vary, according to reports from qualitative illness in the household. These findings confirm that data.) adverse selection is present in the scheme. From a public health perspective, these results are positive in that they What are the most important factors affecting enrollment? indicate that people who most need health care services Although the comparison of the two groups discussed are purchasing insurance. However, adverse selection can above is useful for understanding the sample, it is only drive up the cost of health care per insured member, possible to determine whether a variable is associated with thereby threatening the sustainability and financial viability enrollment when we control for differences in other of the scheme. characteristics. As described under Study Approach, the predicted probabilities shown in Figures 4 and 5 examine According to insurance theory, people who do not like to the relationship between CBHI enrollment and the take risks will be more likely to enroll in health insurance variable of interest (at the household and village level, due to the desire to protect themselves from health- respectively), while holding all other independent related financial loss in the future [24]. However, the variables constant at their mean value. Only variables that study found that people who are very risk averse are are significantly associated with enrollment are presented actually less likely to enroll in CBHI. Qualitative here. (A description of the relationship between interviews shed some light on why this may be the case. enrollment and the full set of variables in the model can be Although the majority of the respondents in the qualitative found in the technical report). interviews reported that enrolling in CBHI allows people to minimize their risk, some felt that enrolling in CBHI is Household level determinants a risky venture and that enrollment actually increases risk (because one can't be sure that benefits will be delivered Age and education of household head. At the household level, when they are needed). age of the household head influences enrollment in CBHI, such that households in which the household head is older Other types of risks at the household level are associated are significantly more likely to enroll. Although with enrollment: having a pregnancy in the household and households with primary or secondary education are no having more women of reproductive age. However, more likely than those with no education to enroll, the having an older family member or children under the age of five (results not presented here) does not significantly 12 Employment excludes unpaid household duties but includes increase the probability of enrollment, when all other subsistence farming. factors are held constant. 13 See Appendix 1 for more details on risk measurement. 5 Figure 4. Predicted probability of enrollment by household characteristics 25% 20% 15% 10% 5% 0% S/o in hh has chronic hlth condition Age of HH Head (30 years) Age of HH Head (50 years) Age of HH Head (65 years) Primary HH consumption per capita - Q1 HH consumption per capita - Q2 HH consumption per capita - Q3 HH consumption per capita - Q4 HH consumption per capita - Q5 1 women of reproductive age 3 women of reproductive age No illness in household No education Secondary education HH head willing to take some risk 4-5 people S/o in hh has worse than avg health Quality perception is less than good Quality perception is good Vocational training HH size: 1-3 people S/o had difficulty w activities in 3 mos Worse than avg + chronic + difficulty Post-secondary education HH size: 6+ people CBHI campaign: never attended Chronic + difficulty HH head is risk averse No pregnancy CBHI campaign: attended Pregnancy Worse than avg health + chronic * Variables represented by light colors are not significantly different from reference group (dotted bars). Figure 5. Predicted probability of enrollment by village characteristics 25% 20% 15% 10% 5% 0% Urban Village chief <=40 Semi-urban or rural <3 campaigns in village 3+ campaigns in village Hatxaifong district Sissatanak district Size of village: 50 Keoudoum district No drug seller Village chief > 40 Drug seller Chief's educ=sec.+ Phonghong district Chief's educ.8 persons 33,000 28,000 Monks, nuns, dormitory students 5,000 5,000 15 Appendix 2. Background characteristics of CBHI and non-CBHI households CBHI Comparison p-value (n=1000) (n=2000) Sociodemographics Mean household size (persons) 5.32 4.71 <0.001** Marital status of household head (% married) 84.17% 80.39% 0.027* Education Highest level=any primary 43.10% 42.70% 0.866 Highest level=any secondary 31.65% 37.23% 0.028* Highest level=university/institute 5.08% 2.30% 0.002** Highest level=vocational 11.82% 8.41% 0.020* Age of HH head (mean years) 52.44 48.44 <0.001** Ethnic majority (1=Kai-Tadai; 0=other) 98.56% 98.16% 0.404 Total annual household consumption ($US) $3,161.97 $2,729.33 <0.001** Total annual per capita consumption, mean ($US) $753.81 $741.12 0.531 % HHs below the national poverty line 19.68% 20.19% 0.757 Employment status Not working for money 21.10% 17.17% 0.009** Family farm-based agriculture 24.05% 22.85% 0.644 Small-scale trading or family business 26.43% 31.21% 0.039* Work for someone else 28.42% 28.77% 0.878 Total (employment) 100% 100% % HH heads with long-term employment contract (12 mos +) 17.19% 11.58% 0.002** Health status and risk aversion % HHs in which average self-rated health is below average (<3 on scale of 1 to 5) 19.43 14.93 0.023* % HHs in which someone has disability or chronic condition 23.45 14.54 <0.001** % HHs in which someone had difficulty performing daily activities in past 3 months 16.29 10.98 0.008** % HHs in which someone has experienced deterioration of health in past year 11.89 8.51 .034* 21 Risk preferences : head of household is risk-averse 37.13 41.6 0.041* Other risk variables % of HHs with any member age 65+ 28.05 21.9 .001** % of HHs with any member age 0-5 37.01 37.65 0.754 Mean # of females 15-49 1.57 1.36 <0.001** % HHs in which a woman has given birth in past 2 years 15.74 13.9 0.261 % of HHs in which a woman is pregnant 4.43 2.32 .004** Attitudes toward sources of care and quality perceptions % respondents recommending a government hospital for an uninsured friend...........) a severe condition/emergency? 99.38 99.54 0.669 a moderate condition? 94.55 92.58 0.138 a minor condition? 97.5 96.82 0.420 % of respondents stating that services at district hospital are good 75.44 64.77 <0.001** Per capita expenditure was calculated using adult equivalents but yields similar results when calculated using equal weights for all household members. There was no empirical basis from Laos on the choice of the adult equivalent parameter. **Difference is significant at the .01 level. *Difference is significant at .05 level. A p-value is a statistical term that measures the likelihood that a difference observed in the sample is due to chance. A result that is significant at the 5% level has a p-value of less than .05. Reported results are based on t-tests of means for continuous variables and chi-squares for proportions/ categorical variables. All estimates account for sampling weights and village-level clustering. 21 This question presents the respondent with a gamble in which he/she must guess which hand contains money. The first pick is the same regardless of the hand selected but in the next bets, the stakes become increasingly higher. The variable was dichotomized to differentiate those who were completely risk averse from those who will take at least some risk. 16 Appendix 3. Exposure to CBHI and Trust in Schemes Exposure to CBHI/Trust CBHI Comparison p-value (n=1000) (n=2000) % attended CBHI campaign 92.52% 66.24% <0.001** How many of your close relatives/friends had joined CBHI prior None 4.49% None 30.74% <0.001** to enrollment? (or how many are enrolled now?) Some 49.18% Some 48.89% Many 46.33% Many 20.37% % of HHs reporting that they trust that the money contributed to 92.54% 69.72% <0.001** CBHI will be used in the right way % of HHs reporting that members will get the benefits they pay 95.84% 69.44% <0.001** for when they need them **denotes statistical significance at the .01 level. Reported results are based on Chi-square tests. All estimates account for sampling weights and village-level clustering. Appendix 4. Quality ratings of district hospital Quality ratings (1 to 5) CBHI Comparison p-value (n=1000) (n=2000) The way staff act towards patients 3.65 3.46 <0.001** The quality of the facilities and equipment 3.67 3.54 0.005** The skills/competence of staff 3.74 3.58 0.001** **denotes statistical significance at the .01 level. Reported results are based on t-tests of means. All estimates account for sampling weights and village-level clustering. Appendix 5. Utilization of health services and out-of-pocket expenditures, for matched observations CBHI Comp % diff T-stat N(CBHI) N(comp) Inpatient (IP) visits IP visit in last year (%) 0.0812 0.0422 92% 8.81** 5298 9468 Number of IP visits in last year (for those with visit) 1.1615 1.1267 3% 0.94 422 368 IP visit at any public facility in last year (%) 0.0809 0.0396 104% 9.31** 5298 9468 IP visit at district hospital in last year (%) 0.0379 0.0091 316% 10.61** 5298 9468 IP visit at prov. hospital in last year (%) 0.0315 0.0172 84% 5.08** 5298 9468 IP visit at central hospital in last year (%) 0.0112 0.0137 -18% -1.05 5298 9468 OOPs on all IP visits in last year (incl. transport and food) (LAK) 610,000 1,600,000 -62% -6.53** 422 368 OOPs per IP visit in last year (incl. transport and food) (LAK) 530,000 1,400,000 -62% -7.15** 5298 9468 Outpatient (OP) visits OP visit in last month (%) 0.1589 0.1329 20% 4.22** 5298 9468 Number of OP visits in last month (for those with visit) 1.23 1.17 5% 1.72 802 1279 OP visit to any public facility in last month (%) 0.068 0.025 173% 10.97** 5298 9468 OP visit to a district hospital in last month (%) 0.047 0.009 416% 12.7** 5298 9468 OP visit to a provincial hospital in last month (%) 0.015 0.008 92% 3.46** 5298 9468 OP visit to a central hospital in last month (%) 0.004 0.005 -17% -0.52 5298 9468 OP visit to a pharmacy in last month (%) 0.086 0.092 -7% -1.3 5298 9468 OP visit to a private clinic in last month (%) 0.007 0.013 -42% -2.78** 5298 9468 OP visit to a healer in Laos in last month (%) 0.000 0.001 -54% -1.02 5298 9468 OOPs on all OP visits in past month (LAK) 8,596 69,598 -88% -3.42** 802 1279 OOPs on all OP visits in public facility in last month (LAK) 8,424 38,981 -78% -3.36** 802 1279 OOPs on all self-care in last month (LAK) (for all individuals) 220 298 -26% -0.65 802 1279 OOPs per OP visit public facility in last month (LAK) 6,374 32,586 -80% -3.08** 802 1279 OOPs per self-care visit in last month (LAK) 651 1,457 -55% -1.74 802 1279 OOPs per individual (all IP and OP costs + premiums) 370,000 1,100,000 -66% -4.00** 1138 1573 Average OOPs for entire sample, regardless of use 130,000 170,000 -24% -1.29 5298 9468 **denotes statistical significance at the .01 level ; * denotes statistical significance at the .05 level 17